Human Computer Interaction, Artificial Intelligence and Intelligent Augmentation
Loading...
Paper Type
short
Paper Number
1368
Description
Algorithmic decision support is omnipresent for many managerial tasks where human judgment makes the final call. However, the lack of transparency of algorithms is often stated as a barrier of successful human-machine collaboration. In this paper, we analyze the effects of algorithm transparency on the perceived value of algorithmic advice and its resulting utilization for a simple, easy-to-understand algorithm. In a laboratory experiment, participants received algorithmic advice for a forecasting task. Only the treatment group was informed about the underlying principles of the simple yet optimal advice-giving algorithm. While the explanation increased the understanding of the algorithmic procedure, it reduced the perceived value of the algorithmic advice, its utilization, and the participants’ performance. Our results indicate that the effects of algorithm transparency on the use of algorithmic advice are not straightforward, and that transparency might even be harmful. Going forward, we plan to explore whether algorithm complexity moderates this effect.
Recommended Citation
Lehmann, Cedric Alexander; Haubitz, Christiane; Fuegener, Andreas; and Thonemann, Ulrich, "Keep It Mystic? – The Effects of Algorithm Transparency on the Use of Advice" (2020). ICIS 2020 Proceedings. 3.
https://aisel.aisnet.org/icis2020/hci_artintel/hci_artintel/3
Keep It Mystic? – The Effects of Algorithm Transparency on the Use of Advice
Algorithmic decision support is omnipresent for many managerial tasks where human judgment makes the final call. However, the lack of transparency of algorithms is often stated as a barrier of successful human-machine collaboration. In this paper, we analyze the effects of algorithm transparency on the perceived value of algorithmic advice and its resulting utilization for a simple, easy-to-understand algorithm. In a laboratory experiment, participants received algorithmic advice for a forecasting task. Only the treatment group was informed about the underlying principles of the simple yet optimal advice-giving algorithm. While the explanation increased the understanding of the algorithmic procedure, it reduced the perceived value of the algorithmic advice, its utilization, and the participants’ performance. Our results indicate that the effects of algorithm transparency on the use of algorithmic advice are not straightforward, and that transparency might even be harmful. Going forward, we plan to explore whether algorithm complexity moderates this effect.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.